Determination of diphenylamine residue in fruit samples using spectrofluorimetry and multivariate analysis

Determination of diphenylamine residue in fruit samples was studied based on normal, synchronous first- and second- derivative spectrofluorimetry. These methods were performed using three of the most widely employed multivariate calibration techniques which are partial least squares, multiple linear regression and principal component regression. Eighteen combinational methods were tested to present the best model for determination of diphenylamine residue. The prediction performance of the calibration models, which was constructed on the basis of these methods, was also compared. For a range of concentrations from 10 to 100 mu g kg(-1) of diphenylamine in the fruit prediction set, the values of root mean square error and relative error of prediction, using multiple linear regressions, were determined in the range of 3.3-4.1 mu g kg(-1) and 6.4-8.0%, respectively. Repeatability studies were satisfactory, giving RSD% values of 1.8, 5.6, and 3.3 for apple, pear, and orange, respectively. The calibration graphs were linear in the range of 10-100 mu g kg(-1) and detection limits were between 4 and 7 mu g kg(-1). The presented method was successfully applied to determine diphenylamine residue in some of naturally treated fruit samples and the results indicated proper agreement with those obtained by HPLC analysis. (C) 2013 Elsevier Ltd. All rights reserved.